Sparse Neural Networks

sellProject sellDFKI sellAugmented Vision

Sparse Neural Networks

Topic

In this project, you will train and evaluate several human pose estimation networks with sparsity constraints and compare their runtime performance on different hardware.

Tasks

  1. Learn about and implement network sparsification methods​.
  2. Train multiple human pose estimation networks​.
  3. Deploy them to ONNX and evaluate runtime performance in different environments (desktop, web, and mobile)​.

Expected Skills

  1. PyTorch + good programming skills (required)​
  2. OpenMMLab libraries, e.g., mmengine, mmpose, mmrazor, mmdeploy, etc. (highly preferred)​
  3. Sparsity in neural networks (preferred)​
  4. Familiar with ONNX and mobile deployment (optional)